Meta is launching a new AI model as the global tech race heats up
Meta on Thursday unveiled a new version of its flagship artificial intelligence model, the Muse Spark, in a bid to close the gap with its rivals in the global race to develop the technology.
The company will offer a paid version of the service for the first time, a departure from its long-standing philosophy of giving away its AI for free. The paid product opens up a new revenue stream for Meta, which plans to spend tens of billions of dollars on AI this year.
The new model is better at coding than the previous version of the Meta technology that the company released in April. In tests that measure typing, reasoning, coding and other tasks, Muse Spark performed at or near the same level as leading models from Anthropic, OpenAI, Google and xAI, according to data shared by the company.
The Meta product would also be cheaper than other flagship models, the company said. The Muse Spark costs customers about a quarter of what Anthropic’s flagship model, the Fable, does.
Muse Spark is the first AI model developed by Meta Superintelligence Labs, the AI division of the company that Mark Zuckerberg, Meta’s CEO, revamped last year. In addition to spending billions on new AI researchers to develop models like the Muse Spark, Mr. Zuckerberg has pledged to spend $600 billion in the coming years to build new data centers.
The new AI division is led by Alexandr Wang, the 29-year-old head of AI at Meta and former CEO of the ScaleAI startup. Meta released its first image generator developed under Mr. Wang, called Muse Image, on Tuesday. Meta also plans to release a video generator called Muse Video and an even more powerful AI model internally called Watermelon in the coming months.
Meta faces stiff competition from rival technology firms in the AI race. Elon Musk’s AI company, xAI, released a new AI model on Wednesday, and OpenAI released its latest model on Thursday.
Mr Zuckerberg touted the new Meta model on X on Thursday, calling it “the most powerful in agent performance, tool use and computing”.
“More coming soon,” he added.